Re: [patch 2/3] scheduler: add full memory barriers upon taskswitch at runqueue lock/unlock

[Steven Rostedt]

[ Added linux-arch ]

On Mon, 2010-02-01 at 17:29 +0100, Peter Zijlstra wrote:
> On Mon, 2010-02-01 at 11:24 -0500, Steven Rostedt wrote:
> > 
> > Doesn't set_bit imply a wmb()? If so couldn't we do:
> 
> Nope, that's what we have smp_mb__{before,after}_set_bit() for.
> 
> On x86 its a LOCK'ed op, so sure it implies a full membarrier, but not
> in generic.
> 
> on x86 we have plenty serializing instructions before and after rq->curr
> is set so none of the crazyness is needed at all. The only thing is !
> x86.
> 

So if we go with Linus's approach and make all archs guarantee that
switch_mm() implies a full mb(), then we can simplify the
sys_membarrier() code?

That looks like the best approach.

-- Steve



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/